When NIFTY actually rises and falls

The clock: where the average move sits2015-01-09 to 2026-04-23

The five-minute chart shows the unconditional average return in each cash-session clock bin. Only two full-history bins survive correction for testing all 75 bins: the opening five minutes and 15:10 to 15:15, both negative.

The latest two years491 sessions

This is the window closest to the question. The first-hour dip survives the pre-specified window tests. Across all 75 five-minute clock bins, only the 15:05 rise and the immediately following 15:10 dip survive BH-FDR; the whole cash-session decline, midday and closing hour do not.

Recent FDR survivors: 15:05-15:10 (+1.55 bp), 15:10-15:15 (-1.28 bp). Adjacent opposite-sign closing bins are a microstructure pattern, not a directional closing-hour edge.

Did the pattern weaken?2024-11-18 to 2026-05-14

The fixed trailing 18-month window is a sensitivity check selected by duration, not a hand-picked breakpoint. It shows whether the opening estimate remains stable in later data.

What the FII file can and cannot say489 daily observations

FII index futures were net short on 412 of 489 available days (84.3%) in the aligned two-year window. That is persistent positioning, but not proof of persistent profits. Among 19 net-short month-end signals with a following-month return, NIFTY rose in 9 and fell in 10. FII stock futures were net long on 100.0% of available observations. A portfolio can therefore look short in index futures and long elsewhere at the same time.

After current net-short observations, the mean 20-session NIFTY return was +84.62 bp, not negative. Its horizon-aware HAC interval [-6.03, +175.28] uses lag 19, and its circular block-bootstrap interval [-4.63, +178.07] uses a 20-session block; both include zero.

The 20-layer audit trailNo hidden leap

1. Provenance. Two locally maintained NIFTY minute stores are named and date-bounded. They are not claimed to be statistically independent or separate market-data vendors.
2. Integrity and cross-store gate. Only 2747 complete 375-minute sessions survived; missing minutes were never filled. Across 1208 overlapping sessions, the two local stores' first-hour returns correlated at 1.000000; this is a consistency check, not independent-source validation.
3. Return definition. All returns are additive log returns in basis points; overnight, cash-session and close-to-close returns remain separate.
4. Full-history decomposition. Overnight averaged +10.80 bp while open-to-close averaged -6.69 bp.
5. Recent two-year check. Open-to-close averaged -2.75 bp and its HAC interval includes zero; the first hour averaged -4.43 bp.
6. Late-window sensitivity. In the fixed trailing 18-month slice, first-hour weakness was -3.28 bp. This is a sensitivity check, not an estimated structural break.
7. One-minute clock. The 375-bin minute profile has its own BH family but remains inspection-only because minute-level microstructure is noisy.
8. Five-minute clock. Full history: 2 of 75 bins survive BH-FDR at 5%; recent two years: 2 of 75.
9. Fifteen-minute clock. 1 of 25 bins survive BH-FDR at 5%.
10. Thirty-minute clock. 1 of 12 bins survive BH-FDR at 5%. Twelve non-overlapping bins cover 09:15-15:15, or 360 of 375 cash-session minutes; the final 15 minutes are omitted rather than mixed into a half-width bin.
11. Weekday split. Weekday rows are exploratory subgroup estimates, not five independent trading rules.
12. Expiry split. The expiry comparison uses a disclosed weekday proxy and is not treated as an official holiday-adjusted expiry calendar.
13. Month-end split. Month-end is defined as the last complete cash session in each observed month.
14. Gap conditioning. Gap-up, gap-down and flat-open subgroups are descriptive; thresholds were not selected on a holdout set.
15. Path dependence. Prior-session direction and the opening gap are lagged before conditioning, avoiding same-session look-ahead.
16. VIX regime. VIX quartiles use the prior available VIX close, not the same-day closing value. Exact ties are broken by dated row order, so equal VIX values can fall into adjacent buckets.
17. Trend regime. Bull and bear regimes use the prior NIFTY close versus its prior 200-session average.
18. FII test. FII index futures were net short in 84.3% of available observations from 2024-05-15 through 2026-05-14, but aligned sign tests do not establish a profitable short-every-month edge.
19. Robustness. Segment and clock means use Newey-West lag 5. Overlapping participant forward returns use lag max(5, horizon-1), so the 20-session test uses lag 19, plus 3000 circular block-bootstrap draws with a 20-session block. BH-FDR is applied separately inside each disclosed clock-profile family, not across segment or subgroup tables.
20. Decision contract. Opening weakness is a context prior, never a standalone short signal; direction still requires price structure, invalidation and independent evidence.

Interpretation guardrailsBefore any trade

• Every open option contract has a matched long and short side; rising OI alone does not identify a new writer.
• Public participant OI is end-of-day category data and does not reveal the intraday aggressor or counterparty.
• A short call may be directional, covered, part of a spread, a hedge or market-making inventory.
• Raw contracts are not comparable across lot-size and product-structure changes without normalization.
• No local source used here contains account-level realised option P&L, so it cannot prove FIIs profit by shorting every month.

Estimation: Newey-West/Bartlett lag 5 for segment, clock and subgroup means; participant forward returns use lag max(5, horizon-1), including lag 19 at 20 sessions; 3000 circular block-bootstrap draws, block 20, seed 20260719; applied to headline segment means and participant forward-return means; Benjamini-Hochberg false discovery rate at q=0.05 is applied separately to five clock-profile families: full-history 1-minute (375), 5-minute (75), 15-minute (25), 30-minute (12), and recent-two-year 5-minute (75). It is not one correction across families and is not applied to the pre-specified segment or exploratory subgroup tables. Twelve non-overlapping 30-minute bins cover 09:15-15:15 (360 of 375 cash-session minutes); 15:15-15:30 is omitted rather than represented by a half-width bin. The overlap check compares two local minute stores. It is a consistency check, not independent-vendor or statistically independent source validation. VIX quartiles use prior available closes. Exact ties are broken by dated row order to retain four labels, so equal VIX values can fall into adjacent buckets. Returns are additive log basis points. The expiry subgroup is a disclosed weekday proxy, not a holiday-adjusted official expiry calendar. This is descriptive research before execution costs and is not investment advice.

Official market-structure sourcesPrimary documents

• NSE participant-wise OI file specification
• NSE derivatives reports archive
• NSE clearing mechanism
• SEBI delta-based FPI position-limit framework
• SEBI equity derivatives profit and loss study, FY22-FY24
• NSE NIFTY lot-size revision, 50 to 25
• NSE weekly-expiry consolidation
• NSE NIFTY lot-size revision, 25 to 75
• NSE NIFTY expiry-day revision to Tuesday
• NSE NIFTY lot-size revision, 75 to 65

FAQ4 reader questions · AEO-eligible

The editorial line, distilled. Schema-marked for AI Overview and reader search.

Use the finding